Re: bug in RDF model theory having to do with rdfs:subClassOf and rdfs:subPropertyOf

From: pat hayes <phayes@ai.uwf.edu>
Subject: Re: bug in RDF model theory having to do with rdfs:subClassOf and rdfs:subPropertyOf
Date: Sun, 22 Sep 2002 18:23:40 -0500

> >I was wrong in my initial speculation that making the semantic constraint for
> >rdfs:subClassOf an iff constraint would not have RDFS-observable
> >consequences.
> >
> >If the semantic constraint for rdfs:subClassOf is changed to an iff
> >constraint then
> >
> >	rdf:type rdfs:domain foo .
> >
> >would entail
> >
> >	rdfs:Resource rdfs:subClassOf foo.
> 
> Ah, most ingenious. OK, but I think this is harmless. It doesn't work 
> for arbitrary domains, and the domain of rdf:type is already required 
> to be the universe. What you have noticed is that trying to restrict 
> that domain to a subset of the universe only makes the universe 
> smaller, in effect. The moral is that one cannot sensibly use RDF to 
> restrict the domain of the rdf:type property. Which is in fact hardly 
> surprising, since to do so would be to change the definition of 
> 'class' and hence of subClassOf.
> 
> Also it depends on exactly how the condition is worded. I'll think 
> about that some more.
> 
> Pat

Well, RDF allows one to do these sorts of things, so they have to be
accommodated.  Similar problems may arise in any version of OWL that allows 
restrictions on rdf:type.

peter

Received on Sunday, 22 September 2002 21:34:57 UTC